Skip to content

VibuAchari/NeuroMold

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🧠 NeuroMold: Bio-Inspired Neuromorphic Control for Injection Molding

NeuroMold is an intelligent, neuromorphic edge computing framework designed to optimize real-time control in plastic injection molding systems using Spiking Neural Networks (SNNs). This project leverages bio-inspired computation to achieve ultra-low power, adaptive control on edge devices — mimicking the brain’s ability to respond dynamically to complex stimuli.

Simulation live at: https://superlative-seahorse-ea5599.netlify.app/dashboard


🧬 Project Motivation

Plastic injection molding industries face major bottlenecks such as:

  • ❌ High defect rates (30%+ rework due to undetected flaws)
  • ⚡ Inefficient energy usage and material waste
  • 🧱 Rigid legacy systems (SCADA, PLC) with no predictive capacity
  • 🧠 Traditional ML models (ANNs) too heavy for real-time, on-device inference

NeuroMold proposes a next-gen solution: biologically inspired SNNs embedded in edge devices for real-time anomaly detection, defect prediction, and closed-loop adaptive control.


🧠 Core Architecture

NeuroMold’s technical pipeline is modular and built for simulation and potential deployment:

🔸 1. Synthetic Sensor Module

Generates realistic time-series sensor data (pressure, temperature, etc.) to simulate factory conditions.

🔸 2. Spike Encoding Layer

Custom encoder transforms continuous data into spike trains using temporal Poisson encoding, designed to work with SNNs.

🔸 3. Spiking Neural Controller (SNN)

A Norse-based LIF (Leaky Integrate-and-Fire) SNN processes spike inputs and generates optimal control signals in an energy-efficient, event-driven fashion.

🔸 4. Injection Molding Simulator

A physics-informed model simulates actuator feedback loops, plastic flow, and thermal states to evaluate controller performance.

🔸 5. Feedback & Evaluation Layer

Quantitatively measures:

  • Response latency
  • Energy use (via spike counts)
  • System stability & accuracy

📁 Project Structure

🔬 Simulation Backend (Python + PyTorch)

Simulations/
├── data/
│   └── synthetic_injection_molding_dataset.csv
├── encoding/
│   └── spike_encoder.py
├── model/
│   └── snn_norse_controller.py
├── simulator/
│   └── injection_molding_simulator.py
├── results/
│   └── *.npy  # Includes outputs from ANN, PID, SNN
├── visualizations/
│   └── output_charts/  # PNG graphs
├── run_pipeline.py
├── requirements.txt
└── NeuroMold_Report.md

🧑‍💻 Frontend Dashboard (Vite + TypeScript + Tailwind)

Dashboard/
├── src/
│   ├── components/
│   ├── charts/
│   ├── context/
│   └── utils/
├── public/
├── index.html
├── tailwind.config.js
└── vite.config.ts

🧠 Why Bio-Inspired?

NeuroMold leverages Spiking Neural Networks, which:

  • Mimic real brain neurons firing only when needed (sparse computation)
  • Encode time-dependent signals efficiently
  • Are ideal for event-driven, real-time factory environments

This makes NeuroMold not just another AI controller — it’s neuromorphic intelligence at the edge.


🔬 Key Features

  • 🧠 SNN-based controller using LIF neurons for adaptive control
  • Energy-efficient inference measured via spike counts
  • 🧪 Simulated real-world testbed for evaluating control strategies
  • 📊 Dashboard visualization of metrics like spike rate, energy vs accuracy
  • 🔄 Modular architecture for future hardware integration (e.g. Intel Loihi)

🚀 Running the Simulation

⚙️ Prerequisites

  • Python 3.9+
  • PyTorch
  • Norse
  • NumPy, Matplotlib

▶️ To Run:

cd Simulations
python run_pipeline.py

📈 Output:

  • .npy arrays with SNN/PID/ANN controller outputs
  • Charts in visualizations/output_charts/

🖥️ Running the Dashboard

cd Dashboard
npm install
npm run dev

This starts the dashboard at http://localhost:5173/ with live graphs for spike activity, controller comparison, and more.


🧪 Citation

@misc{achari2025neuromold,
  author       = {Vibusha S Achari},
  title        = {NeuroMold: Neuromorphic Adaptive Control for Injection Molding},
  year         = {2025},
  howpublished = {\url{https://github.com/VibuAchari/NeuroMold}},
}


🧭 Future Directions

  • Integrating real-time sensor streaming via MQTT
  • Optimizing spike encoding through self-adaptive schemes
  • Porting to neuromorphic hardware (Intel Loihi, SpiNNaker)
  • Filing a patent and industrial pilot deployment

NeuroMold aims to push the frontier of bio-inspired intelligent manufacturing — bringing brains to the factory floor, one spike at a time. ⚡🧠🏭

About

Neuromorphic Edge Intelligence for Adaptive Injection Molding Control Systems

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors